用于高效少样本表格分类的语言模型表征

The web serves as a rich source of structured data in the form of tables, including product catalogs, knowledge bases, and scientific datasets. However, the heterogeneity in the structure and semantics of these tables poses significant challenges in developing a unified method to effectively leverage the information they contain. Large language models (LLMs) are increasingly integral to web infrastructure for tasks such as semantic search. This raises a critical question: can we utilize these already-deployed LLMs to classify structured data in web-native tables (e.g., product catalogs, knowledge base exports, scientific data portals) without the need for specialized models or extensive retraining? This study investigates this question and proposes a lightweight approach. By effectively leveraging existing LLMs, efficient table classification can be achieved even with few samples.